Machine learning based visibility estimation to ensure safer navigation in strait of Istanbul

Uyanık T., Karatuğ Ç., Arslanoğlu Y.

Applied Ocean Research, cilt.112, 2021 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 112
  • Basım Tarihi: 2021
  • Doi Numarası: 10.1016/j.apor.2021.102693
  • Dergi Adı: Applied Ocean Research
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aquatic Science & Fisheries Abstracts (ASFA), Artic & Antarctic Regions, Environment Index, INSPEC, DIALNET
  • Anahtar Kelimeler: Safe navigation, Risk assessment, Istanbul strait, Machine learning, Visibility prediction, STATISTICAL-ANALYSIS, RISK-ASSESSMENT, ACCIDENTS, PROBABILITY, CASUALTIES, DECISION, MODEL
  • İstanbul Teknik Üniversitesi Adresli: Evet

Özet

© 2021Maritime transportation is more preferable day by day with the increase in cargo capacity worldwide. Therefore, the number of voyages take place in the seas is increasing and more intensive maritime traffic occurs in narrow channels. One of the difficult and dense seaways in the world is the Istanbul Strait, which is one of the most important strategical channels that connects the Black Sea and the Marmara Sea. Every year many marine vessels navigate in this route and carry dangerous cargos for human health and environment. Thus, ensuring safe navigation in the Strait is an important subject due to prevent a marine accident, save human health, and protect the environment from any disaster. In this study, meteorological knowledge, one of the important factors affecting safe navigation, is provided from a local weather station. The visibility during the passage in the Strait has estimated with various machine learning methods based on wind speed/direction, humidity, pressure, time indicators. To determine the relationship between the variables more clearly, a correlation matrix was created firstly. Different error metrics have used for accuracy and reliability of the predictions. The results of the estimation process show that the Gradient Boosting method is a more successful method.